I don't have much time to spend on this now myself, but I put together a few preliminary ideas and some diagrams that others may find handy for thinking about the problem, at http://www.jsoftware.com/jwiki/MeteorPuzzlePreliminaries .
Also, since this shoot-out remains based on the old-fashioned premise that run-time and code size are the criteria by which a solution should be measured, I've been tracking the time I put into this as we should emphasize the total time-to-solution for solving problems. So far, not counting the hour or so I just spent putting up the Wiki page, I've spent about 2 1/2 hours on this. On 5/26/07, Raul Miller <[EMAIL PROTECTED]> wrote:
On 5/26/07, John Randall <[EMAIL PROTECTED]> wrote: > Here is my plan. > > For each piece, find all possible placements on the board. Represent > these as bit vectors with 50 elements, each stored in a single 64-bit > integer. A combination of placements is illegal if the bitwise and of > the corresponding integers is nonzero. Do depth-first search on legal > lists of configurations for each piece. That would not work as stated, for me, because I have a 32 bit machine, and b. does not currently include such integer valued ieee floats in its domain (I guess the issue has something to do with properly denormalizing and renormalizing those floats): 12345 (17 b.) 2^49 |domain error However, a pair of integers (8 bytes) would be more compact than my current representation (10 bytes), and almost certainly faster to work with. I'll give this a try, maybe this evening or tomorrow. Thanks, -- Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
-- Devon McCormick, CFA ^me^ at acm. org is my preferred e-mail ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
